Wireless networks are used for a variety of applications. Traditionally, the main application has been voice, delivered over a circuit switched network. The network has since been adapted to deliver circuit switched and packet switched data. With higher speed networks, applications such as video telephony will soon become available.
Within such a framework, a limited amount of functionality exists to tailor the service to the environment. For example, Global System for Mobile Communications (GSM) and Wideband Code Division Multiple Access (WCDMA) networks use the Adaptive Multi-Rate (AMR) vocoder for voice, which facilitates changes in the source rate with changing channel conditions in order to enable the best perceptive performance. Within AMR, means exist to signal to a distant vocoder that receiver conditions have changed and that a different vocoder rate is preferable.
Typically, in such a network, one vocoder is at the mobile device, and another vocoder is at a transcoder unit or media gateway in the network. The media gateway converts the encoded voice into a form that can be transported over other networks. Thus, the rate change for voice communication occurs between the mobile device and the media gateway. Generally, information about the quality of the radio link is available only at the base station and the mobile device, and this information has to be signaled to the vocoder, which is available locally at the mobile device, or is a remote node in case of the media gateway. In GSM networks, in-band signaling within the AMR packets is used to signal the desired vocoder rate to the vocoder at the remote end; thus the vocoder rate can be adjusted depending on the condition of the individual radio link. In WCDMA networks, an out-of-band signaling is used to signal the desired vocoder rate. This is particularly applicable to circuit based voice transmission.
Wireless networks are gradually being converted to packet switched architectures with a trend towards all application data being delivered using packet switching. For example, voice over IP (VoIP) is being considered as the primary means of delivering voice for telephony. Most other applications, including video, audio and other multimedia applications will also be delivered via the packet network. Real time source information for applications such as these is typically sent using Real-Time Protocol (RTP) which attaches a timestamp and a sequence number to every packet at the source. The RTP packet is transmitted using User Datagram Protocol (UDP) and the Internet Protocol (IP). At the destination, the receiver may use the timestamp information to replay the packets at the correct relative time. Additionally, the timestamp can be used to determine whether the packet has lost its usefulness along the path to its destination. The sequence number is used to track lost or duplicated packets. The RTP protocol operates between the source and the receiver. Some sources may send information in a format that is not interpretable by the receiver; the media gateway can then convert this to a format that is understood by the receiver. Such operation is negotiated during the set up of the session. In this case, the RTP protocol operates between the media gateway and the receiver.
In the IP framework, a limited amount of adaptation to the link capability is possible. Many sources encode data in a format that is suitable for the average quality of the link. This is often done by querying the receiver about the link capability (e.g., 128 kbps, 1 Mbps, etc.), or can be done automatically by sounding the link and getting feedback as to the data capability. However, this method is not suitable for links whose quality can vary dynamically. The quality of wireless links varies at the speed of the fading, and can also vary with varying traffic levels. In such cases, the delay in providing feedback may be large enough that efficient adaptation to the link is not possible. In packet switched radio networks, the optimum transmission format (such as vocoder rate) has to take into account the instantaneous radio channel quality, and other factors such as the traffic load and the retransmission schemes (such as hybrid ARQ). Most of this information is only available at the base station, and not at the remote source coder. There are currently significant technical obstacles to sending information on all these parameters to the remote source coder in an expedient manner so that the coder can react to instantaneous conditions.
The so-called M-pipe project is currently being designed to overcome the significant technical obstacles hereinabove described. The M-pipe project is directed to developing new media codecs that produce scalable encodings that can be locally adapted to conditions, and in the development of signaling mechanisms that can be used to convey to network nodes the scalable nature of such encodings. Disadvantageously, the project lacks the ability to use existing media codecs that do not produce scalable media.
Progressive or scalable coding is utilized in systems such as DVB-T and DVB-H, wherein hierarchical modulation schemes (e.g. 16QAM and QPSK) are used to convey source data of different fidelity to users with different radio conditions. Such operation is mainly for broadcast transmission, and no notion of adapting to a particular user is considered in these systems.
In U.S. Pat. No. 7,194,000 to Balachandran and Ramesh a method is described wherein progressive encodings of source data are encoded in multiple packets with different priorities being assigned to the packets based on the importance of the data. The scheduling function is capable of dropping packets of lower priority, but is still able to deliver sufficient source data to the receiver in order to render the multimedia information at some level of fidelity. Disadvantageously, this method only allows a coarse level of adaptation to radio conditions.
What is desired is a method and system adapted to select the best format for transmission without the need for such detailed feedback.